Bivariate correlations between 142 socio-demographic and health status characteristics of clients at baseline and the 4 network measures were used to identify potentially confounding client characteristics to be included as covariates in subsequent multivariate analyses (Table 4).
Next, four sets of repeated measures mixed regression models were used to examine the association of each network measure with 16 client service use measures and 18 client outcome measures during their first year in CICH (Table 5). Prior multivariate analyses showed significant changes over time among these 34 client service use and outcome measures during the first year of program participation (Mares & Rosenheck, 2007; Table 2).
The main effects of time and status on each of the four network measure (independent variables) were examined, covarying for the baseline value of each client service use or client outcome measure (dependent variable) and additional client baseline characteristics bivariately associated with each network measure.
Thus, a total of 136 mixed regression models were used evaluating the relationship between 4 network independent variables and each of 34 client dependent variables (4*34= 136)(Table 5). Statistical significance for network measures are presented in this report, along with coefficients for those network measures found to be significantly associated with client service use and outcomes measures at the p<.05 level of significance.
Significant results at the Bonferroni-adjusted p<.001 were highlighted in bold type. The Bonferroni correction corrects for multiple-comparisons when several statistical tests are being performed. The smaller p<.001 level of significance was calculated by dividing the conventional p<.05 level of significance by 34, the number of dependent variables tested (i.e., .05/34=.001).